From b49a304daee2db08044bffa65f3894c942b5185e Mon Sep 17 00:00:00 2001
From: mehmet yusufoglu <mehmetyusufoglu01@gmail.com>
Date: Fri, 19 Jul 2024 13:04:54 +0200
Subject: [PATCH] some refactoring

---
 example/matrixAddWithMdspan/src/matrixAddMdSpan.cpp |  5 +----
 example/matrixMulWithMdspan/src/matrixMulMdSpan.cpp | 13 +++++++------
 2 files changed, 8 insertions(+), 10 deletions(-)

diff --git a/example/matrixAddWithMdspan/src/matrixAddMdSpan.cpp b/example/matrixAddWithMdspan/src/matrixAddMdSpan.cpp
index e80edfb0b22..eb76941bd7e 100644
--- a/example/matrixAddWithMdspan/src/matrixAddMdSpan.cpp
+++ b/example/matrixAddWithMdspan/src/matrixAddMdSpan.cpp
@@ -130,7 +130,6 @@ auto example(TAccTag const&) -> int
     auto mdDevB = alpaka::experimental::getMdSpan(bufDevB);
     auto mdDevC = alpaka::experimental::getMdSpan(bufDevC);
 
-
     //  Let alpaka calculate good block and grid sizes given our full problem extent.
     auto const workDiv = alpaka::getValidWorkDiv<Acc>(
         devAcc,
@@ -142,11 +141,9 @@ auto example(TAccTag const&) -> int
     // Execute the kernel
     alpaka::exec<Acc>(queue, workDiv, MatrixAddKernel{}, mdDevA, mdDevB, mdDevC);
 
-    // Wait for the kernel to finish
-    alpaka::wait(queue);
-
     // Copy result back to host
     alpaka::memcpy(queue, bufHostC, bufDevC);
+    // This wait is not necessary if the queue is a blocking queue
     alpaka::wait(queue);
 
     // Verify the result
diff --git a/example/matrixMulWithMdspan/src/matrixMulMdSpan.cpp b/example/matrixMulWithMdspan/src/matrixMulMdSpan.cpp
index 96b5b55333d..19b957ec023 100644
--- a/example/matrixMulWithMdspan/src/matrixMulMdSpan.cpp
+++ b/example/matrixMulWithMdspan/src/matrixMulMdSpan.cpp
@@ -44,11 +44,14 @@ struct MatrixMulKernel
     //! \param C Output matrix where the result of A * B will be stored
     //! \param K The shared dimension between A and B
     template<typename TAcc, typename MdSpan>
-    ALPAKA_FN_ACC void operator()(TAcc const& acc, MdSpan A, MdSpan B, MdSpan C, Idx K) const
+    ALPAKA_FN_ACC void operator()(TAcc const& acc, MdSpan A, MdSpan B, MdSpan C) const
     {
-        // compile time check
+        // compile time checks
         static_assert(isMdspan<MdSpan>::value, "The type MdSpan should be an std mdspan");
 
+        // A is MxK and B is KxN
+        auto const K = static_cast<Idx>(A.extent(1));
+
         auto const i = alpaka::getIdx<alpaka::Grid, alpaka::Threads>(acc)[0];
         auto const j = alpaka::getIdx<alpaka::Grid, alpaka::Threads>(acc)[1];
 
@@ -146,13 +149,11 @@ auto example(TAccTag const&) -> int
         alpaka::GridBlockExtentSubDivRestrictions::Unrestricted);
 
     // Execute the kernel
-    alpaka::exec<Acc>(queue, workDiv, MatrixMulKernel{}, mdDevA, mdDevB, mdDevC, K);
-
-    // Wait for the kernel to finish
-    alpaka::wait(queue);
+    alpaka::exec<Acc>(queue, workDiv, MatrixMulKernel{}, mdDevA, mdDevB, mdDevC);
 
     // Copy result back to host
     alpaka::memcpy(queue, bufHostC, bufDevC);
+    // This wait is not necessary if the queue is a blocking queue
     alpaka::wait(queue);
 
     // Verify the result